We describe a new mixed integer nonlinear programming approach for the automated superstructure generation problem in process synthesis engineering, whose target consists in defining and optimizing the superstructure, i.e., the union of all the alternative structures, of a given chemical process cluster. We develop a mixed integer nonlinear formulation for the problem in order to define the links between the chemical units and we implement it with respect to the integrated process of catalytic reforming and light naphtha isomerization in petroleum refinery. The formulation is based on the definition of a graph, whose nodes are the chemical units and whose arcs represent the links between units. The chemical processes are then replaced by their corresponding surrogate models, which represent a systematic way to mathematically represent the input/output relationships of the given processes.